Predicting Changes in Crude Oil Futures Spreads Using High Frequency Order Flows

Abstract

With the increasing financialization of commodities markets over the past two decades and advent of high frequency exchanges, commodities futures prices are no longer solely determined by fundamental supply and demand dynamics. Crude oil futures in particular are of interest since they are the most actively traded commodity futures contract by volume, and the shape of the crude oil futures curve has experienced considerable fluctuations between backwardation and contango in recent years since the shale bust of 2014 - 2015. By using detailed data from the CME Group, this thesis will examine order flows on NYMEX WTI Light Sweet crude oil with two primary objectives in mind. First, to demonstrate positive correlation between changes in calendar spreads on crude oil futures and several predictor variables on both a daily and an intraday basis. Second, to demonstrate positive correlation between flips and changes in the sign of the calendar spread and order imbalance using high-frequency tick data. We find that previous changes in the spot price of crude oil, in addition to prior changes in the spread, are the most significant predictors of the next change in spread. We also find that time-weighted order imbalance calculated at 1-min frequency is an effective predictor of the sign of the next change in futures spread. By being able to predict these shifts in the futures spread, all players in the market will better understand this phenomenon and be better equipped to anticipate broader changes in the market.